7results about How to "Implement anomaly detection" patented technology

The invention discloses a fault detection and classification method for an energy storage battery and a storage medium, and the method comprises the steps: completing the unsupervised anomaly detection through the normal operation data, obtaining the fault features of the abnormal data for the obtained abnormal data, and completing the classification of the features; and after the abnormity is triggered, calling a supervised classification model to give a fault type probability. A computer program used for executing the method is stored in the storage medium. The method has the advantages of simple principle, easiness in implementation, wide application range and the like.

The application relates to the technical field of crack detection, and discloses a cement pouring surface crack detection method and system, which comprises the following steps: a high-frequency sound wave is emitted to a cement pouring surface structure by using an ultrasonic instrument; when the sound wave meets the cement pouring surface, a reflection is generated to form an echo; the echo is captured by a receiver; and whether the cement pouring surface has cracks and the depth of the cracks is identified according to the time difference of the echo. The application continuously captures crack echoes of a preset data detection amount, carries out peak amplitude sorting and difference value calculation on all captured echo segment signals, and further judges whether there is an abnormal echo fluctuation by comparing the fluctuation amplitudes, so that the double judgment mechanism can more effectively exclude invalid scattering and redundant reflection signals caused by rough cement pouring surfaces and other factors, and the risk of misjudgment or false crack alarm is greatly reduced.

The application discloses a lottery ticket device and a running method thereof. The lottery ticket device comprises a shell forming a paper running channel, a transparent plate arranged at an observation window, a camera for image acquisition, a rubber roller device for conveying paper, a ticket feeding sensor and a position sensing assembly for detecting a paper feeding state, a marking device for marking paper feeding, and a controller. The controller performs unified image acquisition and identification processing on a handwritten betting slip, a computer ticket and an instant ticket through the camera, and controls the marking device to perform printing marking or punching marking according to an identification result. The running method comprises paper feeding detection, image acquisition, type determination and corresponding processing flow. Through the adaptive adjustment structure of the paper feeding channel, the single-camera image identification and the collaborative design of the single-module double-marking structure, the application realizes the integrated operation of automatic identification, classification processing and marking of multiple types of tickets, and significantly improves the processing efficiency while simplifying the equipment structure.

The application discloses a Deepfake video detection method based on multi-domain feature region standard score difference, which comprises the following steps: data set division; video frame division and extraction of a region to be detected; construction of a double-branch convolutional neural network; calculation of the RGB feature and NSCT sub-band image of the region to be detected; frequency domain feature obtained by frequency band fusion of the NSCT sub-band image; response of different level texture features obtained through a texture feature extraction module; output features of the space domain and frequency domain feature branches are spliced along the channel dimension, and input into an abnormal feature discrimination module to obtain a tampered region prediction mask; the tampered region prediction mask is subjected to a full connection layer to obtain one-dimensional features, which are spliced with the output features of the texture feature extraction module, and then subjected to a full connection layer and a Softmax activation function to output a binary classification prediction result. The application can better combine the feature information of the space domain and the frequency domain, strengthen the response to the texture feature, discriminate abnormal tampering traces, and improve the generalization ability of the model.

This application provides a cross-domain model training method based on transfer learning, comprising the following steps: A1. Performing sliding window partitioning on the source system log messages and the target system log messages to obtain corresponding source system log sequences and target system log sequences; A2. Performing equal partitioning on the source system log sequences and the target system log sequences to obtain log sequence pairs; A3. Performing parsing and transformation processing on the source system log messages and the target system log messages to obtain log template vectors; A4. Training the model based on the log sequence pairs, the log template vectors, and the total loss function to obtain a trained LSTM model and a hypersphere model. The technical solution involved in this application, by employing a contrastive learning method to compare the similarity between two features in pairs, is beneficial for quantifying the differences between features, which can reduce model training costs and enhance the detection effect of log anomaly detection.

This application relates to the field of device testing, specifically to a method, controller, and testing camera for detecting abnormal insertion of components. The method includes acquiring a first image of the board before insertion and a second image of the board after insertion; performing position correction on the first or second image based on a reference image in the first and second images; comparing the position-corrected first / second image with the uncorrected second / first image to obtain pin images; and determining whether the insertion is abnormal based on the geometric or structural features of the pins in the pin images. The anomaly detection process requires no manual intervention, improving the accuracy and timeliness of the detection results and increasing the efficiency of anomaly detection.